LONDON – A new report commissioned by the Alan Turing Institute details how AI is transforming the financial services industry, but warns that up-to-date regulation is needed to avoid misuse of the technology and mitigate the potential for unintended consequences.

Authored by Seattle University’s Professor Bonnie Buchanan, the aim of the report is to promote responsible adoption of AI technology. The report reveals that AI is already widespread in many areas of the financial services sector. AI and machine learning (ML) are frequently used today in credit card fraud detection, bank customer service chatbots, and robo-advisors which provide financial advice to consumers.

The report also addresses algorithmic trading, which uses AI techniques to trade on the world’s stock markets. The idea is to simultaneously check multiple market conditions and execute trades at the best possible prices. Algorithmic trading also promises to increase accuracy and reduce mistakes (including “fat finger” trades, where human traders press the wrong button, and human errors caused by psychological and emotional conditions).

“Most hedge funds and financial institutions do not openly disclose their AI approaches to trading for proprietary reasons, but it is believed that machine learning and deep learning play an important role in calibrating real-time trading decisions. It also involves neural networks, fuzzy logic and pattern recognition,” says the report.

Robo-trading

What we do know is that high frequency trading (HFT) uses complex AI systems to make extremely fast trading decisions, often thousands or millions of trades per day, and its use is widespread. According to one paper quoted by the report, HFT is responsible for 50-70% of equity market trades, 60% of futures trades and 50% of treasuries.

While the report points out that two-thirds of the top 30 cited papers of HFTs since 2013 show positive market effects, and argues that this type of trading improves market liquidity, the practice still has many sceptics. Many (human) traders are concerned about the “black box” nature of AI systems. There is also “model risk,” the risk that the model doesn’t accurately represent the real world, and the tendency of HFT to mimic other traders without exploring the underlying value of the company whose shares are being traded.

Algorithmic trading can also potentially freeze markets and create instability when trading strategies interact in unforeseen ways, the report says. It has been blamed for the flash crash of the pound sterling during the 2016 Brexit referendum, during which the value of the pound decreased around 6% in just two minutes, as well as several other flash crashes in the preceding years.

report bookcover

So could AI cause the next big stock market crash? To counter that possibility, the latest AI technology is also used to detect rogue AI traders and ensure regulatory compliance.

For example, at the Hong Kong Stock Exchange, AI is used to detect stock market manipulation and market abuses. Their system analyzes historical trading activity to identify patterns and anomalies, including dramatic swings in stock prices and rises in trading volume. The report also cites the Division of Economic and Risk Analysis (DERA) at the US Securities Exchange Commission (SEC), which is using ML to examine massive datasets for potential cases of misconduct, and the Monetary Authority of Singapore, which is using AI/ML to identify suspicious transactions.

To ensure the stability of global markets, regulation of AI systems such as HFTs is obviously required. While there are existing regulations (such as the European MIFID II regulatory framework), this is a rapidly evolving sector.

Faster Than Oversight

“AI and ML are moving faster than policy makers can understand to the extent it is almost outstripping the current legal and regulatory framework. Technology is opaque and fast moving and regulators find it hard to keep pace, for both the cumulative impact and risks of contagion,” says the report.

The same AI technology used in FinTech has also spawned a spin-off field, RegTech, which aims to streamline compliance and regulatory activities. According to the report, the UK’s financial services authority is examining the possibility of making its handbook machine-readable and fully machine-executable, so that systems can implement the rules directly.

The report concludes that AI will inevitably become ubiquitous in finance, but that legal, ethical, economic and social challenges will arise. Many AI techniques remain untested in financial crisis scenarios, and there is a need for more education on AI literacy and awareness.